Rasa chatbot behaves weirdly on docker, works perfectly on local

Rasa Open Source
1

My rasa chatbot works fine on my local machine. My rasa version is:

Rasa Version      :         3.6.15
Minimum Compatible Version: 3.5.0
Rasa SDK Version  :         3.6.2
Python Version    :         3.9.18
Operating System  :         macOS-14.1.1-arm64-arm-64bit

My directory structure is:

.
├── Dockerfile
├── config.yml
├── credentials.yml
├── data
│   ├── nlu.yml
│   ├── rules.yml
│   └── stories.yml
├── domain.yml
├── endpoints.yml
└── models
    └── 20231219-234027-gray-entropy.tar.gz

My docker file is:

# Use an official Rasa image as a base image
FROM python:3.9.18


# Set the working directory
WORKDIR /app/rasa

# Copy the contents of the current directory to the container
COPY . .

RUN pip install rasa==3.6.15

# Specify the command to run when the container starts
CMD ["rasa","run", "--enable-api", "--cors", "*", "--debug"]

But whenever i dockerize my app it shows the log:

DEBUG rasa.core.policies.ted_policy - Failed to load ABCMeta from model storage. Resource ‘train_TEDPolicy3’ doesn’t exist.

DEBUG root - No model server endpoint found after server start.

INFO root - Enabling coroutine debugging. Loop id 187650336839136.

(just few logs which i found confusing)

Mind you, the shell works, but it only gives default fallback responses.

Edit1:

When I try to run the same configuration locally:

conda create -n rasa3 python=3.9.18 ipython

pip install rasa==3.6.15

everything works fine. The rasa shell works as it should. The default response does not get triggered. Is there a problem with my docker python image?

6

@Ameybot where is your requirements.txt file in the docker ?

Check some of my post related to docker and docker compose maybe it will help you.

7

There is no requirements.txt required in my case. Therefore I don’t have it.

8

@Ameybot share your domain.yml or are you using action server to ?

9

@Ameybot

Check this link and mate the necessary changes in your Dockerfile only and build your model.

You need to check rasa train, image for rasha and the cmd command

Let me know

10

I assume you are expecting the model to be copied into the image but Rasa isn’t finding it. Confirm that you have a model in the models directory before building the docker image. You can also use dive to look at the image file system to confirm that the model is in the models directory.